Video motion provides useful cues in understanding video content. As a result, research efforts are increasingly relying on semantic event analysis to obtain video structures and indices. As one of important cues for semantic event analysis, compact and effective motion representation is indispensable for video analysis, especially for sports videos. However, conventional semantic video analysis techniques do not adequately utilize video motion information due to its complexities and the lack of effective motion representations. As a result many video motion events go undetected and unanalyzed.